Supporting multiple floating ips in a cluster

[KlavsKlavsen]

I was thinking it would be VERY nice if this could handle "ALL" public ips assigned to a cluster (so we have HA on them).
f.ex. We have 1 public ip - which lands on traefik (a daemonset on all nodes).. this serves all ingress http/https..
BUT we also have a few services that can't be served through traefik.. so we need an extra public ip for those.

I was thinking that if one could instead give it a "list of floating ips"..
to also work with floating ips on dedicated servers (robot api) - it could be done by f.ex. having a label with ip as value (on loadbalancer service objects) - so if one defines such a label - it should ONLY "sync that ip" to the node THAT loadbalancer-service's pod is on.. WDYT ?

That approach should work for both hcloud and dedicated servers AFAIK - and without such a label on the object - the would do "as it does not" - which means point all floating ips it manages to the same hcloud-server.

[costela]

hi @KlavsKlavsen, thanks for the feedback!

I'm not quite sure I understand what you mean though.
We already support multiple IPs: whatever IPs get assigned to a service, if they can be found via the hclud API, we attach them to the underlying node.
So, leaving aside the robot API part (which is #33), what exact scenario are you trying to solve here?

[KlavsKlavsen]

@costela As I understand it - the code tracks one or more floating ips - so they follow the pod of ONE loadbalancer, right?
I have 2 floating ips.. 1. must point to "any of the nodes running traefik" and 2. must follow my teleport-cluster pod (as the external-ip is assigned to the loadbalancer service that points to that pod.
As I understand it current code, does not support that ?